This paper proposes a multi-source information fusion method for underwater target localization and recognition in highly cluttered environments, addressing the limitations of acoustic-only methods under decoy interference. The key innovations include (1) a novel fusion framework that integrates motion intent estimation with Dempster-Shafer evidence theory, enabling target-decoy discrimination when acoustic features become unreliable, and (2) distributed Kalman filtering that uses only angle-of-arrival measurements from multiple platforms to achieve high-precision trajectory estimation without range data, significantly reducing the constraints on heterogeneous sensor deployment. Experiments illustrate that the proposed algorithm effectively improves target recognition accuracy under decoy interference while enhancing underwater operational capabilities.